Graphing Failed Transaction Count and Percentage
I have a UA BT which is filtering for checkout.jsp . I am trying to graph Failed transaction count + failed transaction percentage and correlate the no. of users impacted.
Failed transaction count:
1) When I graph with aggregation sum the numbers are low and when I change that to count the values are very high. I am guessing count aggregation here returns how many visits are included to collect this data point, is that correct?
2) When I change the aggregation to Maximum, the failed transaction count shows value 0 or 1 with same 1 hr resolution as above. Why would it report as 0 or 1?
3) How can I correlate failed transaction count to no. of impacted users or visits?
Failed transaction percentage:
1) What is the recommended aggregation type for failed transaction percentage? For graphing as well as alerting/incidents.
2) What is the difference between Failed UA transaction percentage vs Failed transaction percentage? The reason I ask is i have same filter for both BTs but when I graph both show different values?
3) When I select the aggregation as count, the value shows as 0 but i see the line go up in y axis. Is that expected or I shouldnt be using count here?
Venkat,
The thing to understand about a BT result measure, taking UA BTs as an example, is that the result measure applies to each UA individually. The Failure Count measure doesn't exactly count anything – it just returns 1 for a failed UA and 0 otherwise.
If you graph 'count' for the Failure Count measure, all you're doing is counting the number of UAs in a given period - the number of '0's and '1's. Also, since all the values are either 0 or 1, the maximum and minimum can only be 0 or 1. What you want is 'sum' which will sum up all the '1's and '0's.
For the Failure Percentage measure, you want to chart the average, maximum or minimum. I believe Failure Percentage is (internally) a rate measure, which means that the values are calculated depending on the time period (i.e., chart resolution or alerting time period).
If you're thinking about alerting, you might consider an alert which uses both measures. That is, generate an alert if (either) the number of alerts over the last five minutes is more than 50 (or) the average failure percentage is more than 10%.
The difference between Failed UA Percentage and Failed Transaction Percentage is that one measures UA failures and the other measures individual transaction (i.e., PurePath, typically web request) failures. You can see both of these configured in the Error Detection pane of the system profile.
-- Graeme
Thanks again Graeme.
Venkat,
If you look at any default installation, you'll see that there are some errors that are classified as "count the transaction as failed (but don't count the UA as failed)" – such as any 4xx errors. That's what will give a failed transaction without a failed UA. You can certainly debate these – shouldn't a 404 mean the UA has failed – but on the other hand most browsers request a favicon.ico file that generates a 404 error and that doesn't really constitute a UA failure.
But it really does depend on the exact configuration you have set up in the Error Detection pane.
-- Graeme
Thanks a lot Graeme. Thats really what I am looking for.
Another question, I see high failed transaction count but nothing on Failed UA count for checkout (unfortunately this is from April and I dont have PP data) . I am not sure whats the exact logic on the failed transaction count. would any exception which is not user impacted could cause this?
You are right. I am looking at these two measurements. "Failed User Action Count" and "Failed User Action Percentage".
When I change the aggregation of "Failed user action count" from sum to count the value is really high. Why is that?
When I change the aggregation of "Failed user action count" from sum to maximum the values are either 0 or 1. ?????
Venkat,
Are you trying to graph one of the UA BT result measures? I'm not sure which measures you're referring to, since the default measures for a UA BT include the UA failure count and percentage, not the transaction failure count and percentage.
Also, assuming you're trying to chart a BT result measure, I find it useful to look at the BT in the BT dashboard and drill down from there to the UA (or PP, etc.) dashboard to see exactly which UAs (or PPs, etc.) are being captured by the BT, and what failure rate is reported there.
-- Graeme
